Ordered rules extraction for incomplete ordered decision system in granular computing

Abstract

Granular computing is a new mathematic analysis method which deals with uncertain information, and it mainly solves problems from different information granularity layers. Aiming at incomplete ordered decision systems, this paper based on granular computing presents a new ordered rules extraction algorithm. Firstly, in order to effectively deal with the incomplete ordered decision system, we transform the incomplete ordered decision system into an extended order value decision table by defining the concept of extended order relation. Then, using the theory of granular computing, we introduce the definition of granular statement, lambda-rank granular statement and lambda-rank granular base in the extended order value decision table. Furthermore, with the search criteria for lowest limit of rule coverage and confidence satisfying user expectation, we design a new algorithm by analyzing the extended order value decision table and granular base from different granularity layers. The algorithm attempts to extract the ordered decision rules as more as possible from granular base in lower rank. Last, we give an application example for proving the validity of the algorithm.